Developing New Computing Approach to Materials Science

As he sees it, managing computing tools to discover new materials involves harnessing the key characteristics of data: volume, velocity, variety and veracity (the four V’s).

Lately, though, “the focus is only on volume,” said Rajan, Iowa State’s Wilkinson Professor of Interdisciplinary Engineering, director of the university’s Institute for Combinatorial Discovery and director of the international Combinatorial Sciences and Materials Informatics Collaboratory. Rajan is also an associate of the U.S. Department of Energy’s Ames Laboratory. “The focus is on more and more data. Data doesn’t make you smarter. What you want is knowledge.”

And so Rajan’s research team is developing statistical learning techniques to research and develop new materials. A 2012 news story in Science by Robert F. Service also contrasts Rajan’s approach with studies that have computed the properties of tens of thousands of potential new battery materials.

“Our approach requires the need to carefully establish a dataset of descriptors on which we directly apply statistical learning tools,” says the Proceedings paper. “One of the arguments we are trying to put forward in this paper is that although the potential number of variables can in fact be large, data dimensionality reduction and information theoretic techniques can help reduce it to a manageable number.”

Rajan likens the process to cooking the perfect spaghetti sauce. Rather than starting with every ingredient in the grocery store, why not start with the most important ingredients? Maybe with the tomatoes and the salt?

“Then how much salt and how many tomatoes?” Rajan said. “Depending on how they’re combined, you get different results. That’s the logic of this.”

The way to start, Rajan said, is to develop some rules of thumb about the material you’re trying to build. Once the most important design rules are set, computing power can be used to search through libraries of compounds and identify promising solutions.

“It’s not that we need more data,” Rajan said. “We need the right data.”

Rajan calls his approach efficient, robust and effective. He says it’s all based on data mining, information theory and statistical learning concepts. He also says it can be readily applied to different problems in various disciplines.

Rajan has used his ideas to help Iowa State researchers advance their work in agronomy, biofuels, climate studies and genomics. His work has been supported by the National Science Foundation, the Department of Defense and Iowa State University.; Source: Iowa State University